Agentic Development

소개

Based on real-world experience building massive codebases entirely with AI, this skill provides a pragmatic framework for high-volume agentic development. It moves beyond theoretical abstractions to offer actionable advice on managing parallel agents in a single folder, leveraging visual context via screenshots, and choosing between CLIs and MCPs based on context efficiency. By treating agents as capable engineers rather than rigid tools, this skill helps developers build the intuition needed to manage blast radius, steer models through strategic interruptions, and maintain high code quality through organic instruction files and same-context testing.

주요 기능

• Strategic refactoring patterns for delegating low-focus mechanical work to agents
• 6 GitHub stars
• Model-specific prompting strategies optimized for both GPT and Claude architectures
• Blast radius estimation for predicting and managing multi-file impact
• Parallel agent management for running multiple concurrent tasks in one directory
• Visual context engineering techniques to reduce token usage and improve accuracy

사용 사례

• Setting up high-velocity development workflows for building large-scale applications with AI
• Optimizing agent output quality through better context management and steering techniques
• Transitioning from rigid development frameworks to intuitive, conversation-driven agent orchestration